JP4243626B2 - Image processing apparatus, image processing method, and computer-readable recording medium - Google Patents

Description

  The present invention relates to an image processing apparatus, an image processing method, and a recording medium on which a program for realizing the image processing method is recorded.

  Conventionally, when transmitting and storing a color image having RGB components, in order to reduce the amount of data, the data is generally encoded (quantized) and compressed.

  Here, since the RGB component has a high correlation between components, the compression is performed after the RGB component is color-converted into a lightness signal and a color signal having a low correlation.

  Various methods have been conventionally considered for the color conversion. For example, YCbCr conversion for converting RGB components into lightness signals Y and color signals Cb and Cr by the following conversion matrix is known.

また、より簡易で高速な色変換方法として、ＲＧＢ成分を以下の式により明度信号Ｙと色信号Ｕ，Ｖに変換するＲＣＴ（Reversible Component Transform）が知られている。

As a simpler and faster color conversion method, RCT (Reversible Component Transform) is known which converts RGB components into lightness signal Y and color signals U and V by the following equations.

なお、上式により示された変換は明らかに可逆な変換であるため、明度信号Ｙ及び色信号Ｕ，Ｖに逆変換を施すことによって、明度信号Ｙ及び色信号Ｕ，ＶからＲＧＢ成分を復元することが可能とされる。

In addition, since the conversion shown by the above equation is obviously reversible, the RGB components are restored from the lightness signal Y and the color signals U and V by performing inverse conversion on the lightness signal Y and the color signals U and V. It is possible to do.

  At present, there is a demand for a technique for generating an image with a small error between the color of the image obtained by the restoration and the color of the original image.

  The present invention has been made to meet the above-described needs, and provides an image processing apparatus, an image processing method, and an image processing method for generating an image with little error from an original image from a compressed image signal. It is an object of the present invention to provide a recording medium on which a program for realizing is recorded.

  The above object is an image processing apparatus for processing a color component extracted from an image signal, wherein the color difference per unit error of the color component generated when the color component is quantized is determined for each range in which the color difference is not more than a predetermined value. This is achieved by providing an image processing apparatus comprising color component quantization means for quantizing a color component by dividing component values into a plurality of ranges. According to such means, it is possible to suppress color errors caused by quantizing the color components.

  Here, the color component quantization means can quantize the low frequency component of the color component.

  Another object of the present invention is an image processing apparatus for processing a color component extracted from an image signal, wherein the color component having a maximum color difference per unit error of the color component generated when the color component is quantized is obtained. This is achieved by providing an image processing apparatus including color component quantization means for quantizing color components under different quantization level numbers before and after the value. According to such means, the number of quantization levels is determined in accordance with the color difference characteristic with respect to the color component, so that the quantization efficiency can be increased.

  Another object of the present invention is an image processing apparatus for processing a color component extracted from an image signal, wherein the color component has different quantization levels before and after the value of the color component in which the polarity of the color component changes. This is achieved by providing an image processing apparatus characterized by comprising color component quantization means for quantizing below. Also by such means, the number of quantization levels is determined according to the characteristics of the color difference with respect to the color components, so that the quantization efficiency can be increased.

  Here, the color component quantization means obtains the color difference per unit error of the color component as an average value when the G component is used as a parameter among the RGB components of the image signal, or when the G component is used as a parameter. If it is determined as a set of maximum values of color differences in, color errors with respect to the original image can be effectively suppressed.

  Also, the color component quantization means uses at least one of the color component value that maximizes the color difference or the color component value that changes the polarity of the color component as a threshold value, and at least one of the values before and after the threshold value. If the color components included in the range are linearly quantized, the quantization can be realized with a simple configuration.

  Another object of the present invention is an image processing apparatus for processing a plurality of color components extracted from an image signal, wherein at least one color component is determined according to the values of other color components not to be quantized. This is achieved by providing an image processing apparatus comprising a color component quantization means for quantization. According to such means, it is possible to realize quantization in consideration of the color error characteristic according to the color component.

  Another object of the present invention is an image processing apparatus for processing a plurality of color components extracted from an image signal, wherein color components not to be quantized and color components to be quantized are set as independent variables. The coordinate space is divided into a plurality of subspaces according to the locus formed in the coordinate space by the maximum color difference per unit error of the color component to be quantized, and the color components according to the subspace to which the color components belong This is achieved by providing an image processing apparatus characterized by including a color component quantization means for quantizing. According to such means, the color component can be quantized under a different number of quantization levels or quantization threshold value for each partial space.

  Here, if the color component quantization means linearly quantizes the color component included in at least one partial space, the configuration can be simplified.

  Another object of the present invention is an image processing apparatus for processing first and second color components extracted from an image signal, in a coordinate space having the first and second color components as independent variables. This is achieved by providing an image processing apparatus characterized in that the distance from the locus in the coordinate space of the point where the first and second color components are equal to each other and the first color component are quantized. Also by such means, the first and second color components can be quantized.

  Another object of the present invention is an image processing apparatus for processing first and second color components extracted from an image signal, and the difference between the first color component and the second color component, and at least the first color component. This is achieved by providing an image processing apparatus characterized by quantizing one of the one or second color components. According to such means, the first and second color components can be easily quantized.

  Another object of the present invention is an image processing method for processing a color component extracted from an image signal, wherein the color difference per unit error of the color component generated when the color component is quantized is a predetermined value or less. This is achieved by providing an image processing method characterized by having a step of quantizing the color component by dividing the value of the color component into a plurality of ranges each time. According to such means, a color error caused by quantizing the color component can be easily suppressed.

  Another object of the present invention is an image processing method for processing a color component extracted from an image signal, in which a color component having a maximum color difference per unit error of the color component generated when the color component is quantized. This is achieved by providing an image processing method characterized by having a quantization step for quantizing a color component under a different number of quantization levels before and after the threshold value. According to such means, since the number of quantization levels is determined according to the color difference characteristic with respect to the color component, the quantization efficiency can be easily increased.

  Another object of the present invention is an image processing method for processing a color component extracted from an image signal, wherein the color component has different quantization levels before and after the threshold of the color component where the polarity of the color component changes. This is achieved by providing an image processing method characterized by having a quantization step of quantizing under a number. Also by such means, the number of quantization levels is determined according to the color difference characteristics with respect to the color components, so that the quantization efficiency can be easily increased.

  Here, the color difference per unit error of the color component is obtained as an average value when the G component of the RGB components of the image signal is used as a parameter, or when the G component of the RGB component is used as a parameter. If it is determined as a set of maximum values of color differences, color errors with respect to the original image can be effectively suppressed.

  Further, in the quantization step, if the color component included in at least one range before and after the threshold value is linearly quantized, high-speed quantization of the color component can be realized.

  Another object of the present invention is an image processing method for processing a plurality of color components extracted from an image signal, wherein at least one color component is determined according to the values of other color components not to be quantized. This is achieved by providing an image processing method characterized by having a quantization step. According to such means, it is possible to easily realize quantization in consideration of color error characteristics according to color components.

  Another object of the present invention is an image processing method for processing a plurality of color components extracted from an image signal, wherein a color component not to be quantized and a color component to be quantized are set as independent variables. When the coordinate space is divided into multiple subspaces according to the trajectory formed in the coordinate space by the maximum color difference per unit error of the color component to be quantized, the color according to the subspace to which the color component belongs This is achieved by providing an image processing method characterized by having a quantization step for quantizing components. According to such means, the color component can be quantized under a different quantization level number or quantization threshold value for each partial space.

  Here, in the quantization step, if the color component included in at least one partial space is linearly quantized, the quantization of the color component can be accelerated.

  Another object of the present invention is an image processing method for processing first and second color components extracted from an image signal, in a coordinate space having the first and second color components as independent variables. It is achieved by providing an image processing method comprising the step of quantizing the distance from the locus in the coordinate space of the point where the first and second color components are equal to each other and the first color component . Also by such means, the first and second color components can be easily quantized.

  Another object of the present invention is an image processing method for processing first and second color components extracted from an image signal, the difference between the first color component and the second color component, and at least the first color component. This is achieved by providing an image processing method comprising the step of quantizing one of the color components of either one or the second. According to such means, the first and second color components can be easily quantized.

  Another object of the present invention is a computer-readable recording medium that records a program for processing a color component extracted from an image signal by a computer. The program is used when quantizing a color component. Provided is a computer-readable recording medium characterized in that a color component is quantized by dividing the color component value into a plurality of ranges for each range in which the color difference per unit error of the generated color component is a predetermined value or less. Is achieved. According to such means, an image in which a color error caused by quantizing the color component is suppressed can be easily generated by a computer.

  Another object of the present invention is a computer-readable recording medium that records a program for processing a color component extracted from an image signal by a computer. The program is used when quantizing a color component. Provided is a computer-readable recording medium characterized by quantizing a color component under a different number of quantization levels before and after a color component threshold value that maximizes the color difference per unit error of the generated color component Is achieved. According to such a means, the number of quantization levels corresponding to the color difference characteristic with respect to the color component is determined by the computer, and the color component is quantized under the quantization level number. Can be increased.

  Another object of the present invention is a computer-readable recording medium recording a program for processing a plurality of color components extracted from an image signal by a computer, wherein the program stores at least one color component. This is achieved by providing a computer-readable recording medium characterized by performing quantization according to the values of other color components not to be quantized. According to such means, it is possible to easily realize the quantization of the color component in consideration of the color error characteristic due to the color component by the computer.

  As described above, if the color component is quantized by dividing the color component value into a plurality of ranges for each range where the color difference per unit error of the color component generated when quantizing the color component is equal to or less than a predetermined value, Since color errors caused by quantizing the color components can be suppressed, an image more faithful to the original image can be reproduced.

  In addition, the color component is quantized differently before and after the value of the color component that maximizes the color difference per unit error of the color component that occurs when the color component is quantized, or the value of the color component that changes the polarity of the color component. If quantization is performed under the number of quantization levels, the quantization efficiency can be increased, and therefore image processing can be simplified in both hardware and software.

  Here, the color difference per unit error of the color component is obtained as an average value when the G component of RGB components is used as a parameter, or as a set of maximum values of color differences when the G component is used as a parameter. Then, since the color error with respect to the original image can be effectively suppressed, highly reliable image processing can be realized.

  In addition, a color component value that maximizes the color difference or a color component value that changes the polarity of the color component is set as a threshold value, and the color component is included in at least one range before and after the threshold value. Is linearly quantized, the quantization can be realized with a simple configuration, and the image processing apparatus can be downsized.

  In addition, if at least one color component is quantized according to the values of other color components that are not to be quantized, it is possible to realize quantization in consideration of the color error characteristics according to the color component. Thus, it is possible to generate an image in which color errors are reliably suppressed with respect to the original image.

  In addition, a coordinate space in which the color component not to be quantized and the color component to be quantized are independent variables, and the maximum color difference per unit error of the color component to be quantized is displayed in the coordinate space. If the color component is quantized according to the subspace to which the color component belongs by dividing it into multiple subspaces according to the locus to be formed, the color component can be reduced under a different quantization level number or quantization threshold for each subspace. Since it can be quantized, it is possible to increase the quantization efficiency and simplify the image processing.

  Here, if the color components included in at least one partial space are linearly quantized, high-speed image processing can be realized with a simple configuration.

  In the coordinate space having the first and second color components as independent variables, the distance from the locus in the coordinate space of the point where the first and second color components are equal to each other and the first color component are quantized. Or by easily quantizing the difference between the first color component and the second color component and at least one of the first or second color component. Can be quantized.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. In the drawings, the same reference numerals indicate the same or corresponding parts.

  As an example, consider a case where the color signals U and V obtained after RCT are quantized linearly and positively and negatively without considering the characteristics of the color signals U and V.

  The quantization method described below more specifically averages the color components for each pixel when the color components (color signals) extracted from the image signal are decomposed for each frequency by wavelet transform or the like. It can apply to the low frequency component obtained by this.

  When the RGB components take values from 0 to 255, the color signals U and V both take values from −255 to +255 (9 bits), but the color signals U and V each have 8 bits (256 levels). For example, linear quantization as shown in Table 1 below can be achieved.

ここで、量子化後の色信号Ｕ，Ｖに逆ＲＣＴを施せば、復号されたＲ’Ｇ’Ｂ’成分が得られるが、当然ながら元のＲＧＢ成分とＲ’Ｇ’Ｂ’成分との間には誤差が生じる。そして、かかる誤差は人の目によって色の誤差（例えば公知のＬａｂ空間における色差、以下「Ｌａｂ色差」と呼ぶ。）として知覚される。

Here, if the inverse RCT is applied to the quantized color signals U and V, a decoded R′G′B ′ component is obtained. Naturally, the original RGB component and the R′G′B ′ component An error occurs between them. Such an error is perceived by human eyes as a color error (for example, a color difference in a known Lab space, hereinafter referred to as “Lab color difference”).

  At this time, as shown in Table 1, when the degree of quantization (compression rate) is small, the error of the color is also small and often not perceivable by the human eye. There is a problem that the error reaches a level that can be perceived by human eyes.

  The following specifically describes an image processing apparatus and an image processing method that can obtain a restored image in which a color faithful to the color of the original image is reproduced even when the color signal is quantized at a high compression rate. explain.

[Embodiment 1]
FIG. 1 is a diagram showing the configuration and operation of an image processing apparatus according to Embodiment 1 of the present invention. As shown in FIG. 1, the image processing apparatus according to the first embodiment includes a hard disk 1, a data bus BS, a personal computer PC, and a printer 13. The hard disk 1 includes a first storage area 3, and the personal computer PC includes a random access memory (RAM) 5 having first and second storage areas 7 and 9, and a central processing unit (CPU) 11. The printer 13 includes a central processing unit (CPU) 15 and a random access memory (RAM) 17 having first and second storage areas 19 and 21.

  Here, the hard disk 1, the RAM 5, the CPU 11, and the printer 13 are each connected to the data bus BS. The first and second storage areas 19 and 21 included in the RAM 17 are both connected to the CPU 15.

  In the image processing apparatus having the above configuration, the original image stored in the first storage area 3 of the hard disk 1 is read into the first storage area 7 in the RAM 5 by the CPU 11 ((1)). The CPU 11 partially reads the image read into the first storage area 7, quantizes the image signal by a quantization method described below, and compresses the image signal ((2)). Further, the CPU 11 writes the compressed image signal in the second storage area 9 of the RAM 5 ((3)).

  Thereafter, the CPU 11 records the image signal written in the second storage area 9 of the RAM 5 in the first storage area 19 of the printer 13 ((4)). As a result, the CPU 15 included in the printer 13 decompresses the image by reading the compressed image signal recorded in the first storage area 19 and tracing the quantization method in reverse to obtain a decoded value (( 5)). Then, the CPU 15 writes the decompressed image signal in the second storage area 21 of the RAM 17 ((6)).

  Accordingly, the printer 13 prints out the image according to a predetermined procedure in accordance with the image signal written in the second storage area 21 of the RAM 17. The amount of data to be transmitted to the printer 13 is reduced by the operation as described above, so that the transmission time is shortened, and high-speed printing is possible even in consideration of the time required for the compression and expansion.

  Next, another operation of the image processing apparatus shown in FIG. 1 will be described with reference to FIG. The original image stored in the first storage area 3 of the hard disk 1 is read by the CPU 11 into the first storage area 7 in the RAM 5 ((1)). Then, the CPU 11 partially reads the image signal stored in the first storage area 7 and quantizes and compresses the image signal by the quantization method described below ((2)). Thereafter, the CPU 11 writes the compressed image signal in the second storage area 9 in the RAM 5 ((3)).

  When all the original images are compressed by the operation as described above, the CPU 11 records the compressed data in the second storage area 4 in the hard disk 1 ((4)).

  Hereinafter, the quantization method according to the present embodiment will be specifically described. Here, since the signal conversion is performed in order to reduce the correlation between the RGB signals as described above, it is desirable to consider the characteristics generated in each component in the RCT result when the signal is compressed. In particular, the characteristics of the color signal component may not be considered as described above when the compression rate is low. However, if the compression rate is to be increased or if it is desired to obtain a high image quality after decompression, this characteristic is required. It is necessary to consider.

  Therefore, in the quantization method according to the present embodiment, for example, it is difficult to perceive a color difference by observation at a predetermined distance, that is, a visual color difference, in other words, color engineering. The color signals U and V are quantized so that an appropriate color difference (for example, a known Lab color difference or a known LUV color difference) is equal to or less than a predetermined value. For example, in the case of RCT, the Lab color difference generated per unit quantization error of the color signals U and V is obtained, and the quantization level necessary for the color signals U and V is set so that the Lab color difference is equal to or less than a predetermined value. The number is determined. The “quantization error” is an error of a color signal caused by quantization, that is, a value obtained when a quantized color signal value is decoded (expanded) is a value before quantization ( This is the error it has with respect to the original value.

  Here, the Lab color difference (Lab color difference / ΔU or Lab color difference / ΔV) generated per unit quantization error of the color signal U or the color signal V changes according to the value of the G component of the image signal. Therefore, for example, as shown in FIG. 3, an envelope 23 obtained by connecting the maximum values when the color signal V corresponding to the value of each G component is used as a variable is used in the quantization described below. It is also good.

  In the quantization, the average value of the Lab color difference obtained for all G component values may be used. As shown in FIG. 4, the G component has a main value such as 0, 50, or 100. A curve 25 obtained by averaging the Lab color difference when taking the image may be used.

  FIG. 5 is a diagram illustrating the color difference caused by the quantization error of the color signal V. In FIG. FIG. 5A is a contour diagram showing the average value of the Lab color difference generated per unit error of the color signal V calculated for the main G component values on the color signal U and V planes. A curve 32 shown in FIG. 5 (b) shows a cross section of the contour map shown in FIG. 5 (a).

  Hereinafter, the method of obtaining FIG. 5 will be described using RCT as an example. In RCT, the color signal V is obtained as a difference obtained by subtracting the G component from the B component. Here, it is assumed that the G component is first fixed and the color signal V is changed by ΔV. That is, it is assumed that the B component of the RGB components increases by ΔV after decoding due to the quantization error.

  At this time, after appropriately setting the color space (so-called input profile) of the RGB signal to be, for example, a standard RGB signal space, the RGB signal is converted by a known method, whereby the coordinate value in the Lab space is converted. Obtainable. Similarly, by converting the RGB component in which the B component is increased by ΔV into the coordinate value of the L′ a′b ′ space, the Lab color difference generated between the original value and the decoded value can be obtained. .

  Here, if the R component and the B component are changed while the G component is fixed, the value of the Lab color difference generated per quantization error ΔV of the color signal V in the UV plane can be obtained.

  From the above, it is clear that the color difference changes using the value of the G component as a parameter, and it is obvious that the change of the Lab color difference using the G component as a parameter can be obtained by a similar method even in conversions other than RCT. is there. Also, the Lab color difference caused by the quantization error of the color signal U obtained as a difference obtained by subtracting the G component from the R component can be obtained by the same method as described above.

  Here, as shown in FIG. 5A, the average value of the (Lab color difference / ΔV) is the smallest in the region 27, and the average value of the (Lab color difference / ΔV) is sequentially increased from the region 28 to the region 31 below. Become. In the region 26, the color signals U and V do not exist.

  In the graph shown in FIG. 5B, the area represents the Lab color difference. Therefore, in the quantization method according to the present embodiment, as shown in FIG. 5B, the area under the curve 32 is a predetermined value in the range where the value of the color signal V takes from −255 to +255. The value of the color signal V is divided into a plurality of ranges (section A to section E) as follows. Then, the number of the divided sections is set as the number of quantization levels, and the quantization threshold value of the color signal V is determined. For example, the decoded value of the color signal V in each section is an intermediate value in the section.

  Further, as shown in FIG. 5, the Lab color difference has an asymmetry with respect to the positive / negative of the color signal V. However, if the color signal is quantized by the above-described method, accurate quantization is performed in consideration of the asymmetry. And the difference in color when viewed with human eyes can be reduced.

  Here, for example, in the above quantization, when the number of required quantization levels is 23, the interval of the color signal V from −255 to +255 is quantized by data having a code length of 5 bits. As described above, the range of the color signal is divided in consideration of the characteristics of the color signals U and V, and the quantization threshold value is set corresponding to the section obtained by the division. If set, the data can be reduced to 5 bits or less while the Lab color difference is set to a predetermined value or less.

  That is, for example, as shown in FIG. 5, when the color difference between the positive and negative color signals V having the same absolute value has a ratio of about 2 to 1 regardless of the value of the color signal U, 0 of the color signal V is obtained. The range in the above is divided into 16, and the negative range less than 0 is divided into 8. The negative range is quantized by a total of 4 bits including 1 bit as a code for identifying positive and negative, and the range of 0 or more includes a total of 5 including 1 bit as a code for identifying positive and negative. Quantized by bit data.

  Therefore, in this case, the color signal can be quantized with data having an average of 4.5 bits while the Lab color difference is kept below a predetermined value. That is, in consideration of the characteristics of the Lab color difference, the number of quantization levels is changed before and after the value (0 in the above example) of the color signal V that maximizes the Lab color difference generated per unit quantization error of the color signal V. As a result, the quantization efficiency can be improved.

  Note that the Lab color difference per unit quantization error may be maximized at the point where the polarity (positive / negative) of the color signal changes, so that the threshold value for changing the number of quantization levels is set as the color signal. The polarity change point of

  Further, as described above, in the section of the color signal V from −255 to +255, in general, nonlinear quantization is performed by using a table storing quantization thresholds. When the color signal V is in a negative range, a non-linear quantization is performed with 4-bit data including a sign for identifying positive and negative, and when the color signal V is 0 or more, a code for identifying positive and negative is used. It is also possible to perform linear quantization with 5-bit data. That is, when quantization is performed by such a method, it is possible to quantize the color signal V of 0 or more without referring to the table. Here, since the table is generally stored in a memory, the amount of memory required at the time of quantization can be reduced by linearly quantizing zero or more color signals V.

  In addition, the linear quantization as described above is employed in the section in which the Lab color difference per unit quantization error of the color signal can be regarded as a constant value, so that quantization with the Lab color difference equal to or less than a predetermined value can be easily realized. Is done.

  In the Lab color difference, the characteristic for the color signal U and the characteristic for the color signal V are different. FIG. 6 is a contour diagram corresponding to FIG. 5A showing the level of color difference caused by the quantization error of the color signal U. FIG. FIG. 6 is a contour map showing the average value of the Lab color difference per unit quantization error obtained for the main G component values on the UV plane, as in FIG. A cross section of the contour map is shown by the graph in FIG.

  Here, the Lab color difference per unit quantization error shown in FIG. 7 depends on the value of the color signal V, but the interval of the color signal U from −255 to +255 is identified as positive or negative in the negative range. Quantization is performed with a total of 4 bits of data including 1 bit as a code, and the range of 0 or more is quantized with a total of 5 bits of data including 1 bit as a code for identifying positive and negative. To improve.

  Here, as with the color signal V, the color signal U is also nonlinearly quantized with 4-bit data including a sign for identifying positive and negative in the negative range, and positive and negative in the range of 0 or more. Linear quantization can also be performed using 5-bit data including a code.

  That is, the quantization method according to the present embodiment is effective for the quantization of the color signal U and the color signal V, and can be said to be a particularly suitable method in application to RCT.

  Further, when comparing the contour maps shown in FIGS. 5A and 6 (the vertical axes in both figures are the same scale), the Lab color difference caused by the quantization error of the color signal V is equal to that of the color signal U. It can be seen that the overall value is larger than the Lab color difference caused by the quantization error. Here, as described above, in both graphs, the area on the graph means the Lab color difference, so that the area is constant between the quantization of the color signal U and the quantization of the color signal V. In order to achieve this, it is effective to change the number of quantization levels.

  That is, in the quantization of the color signal V, the range of the color signal V is divided into a large number of short sections to increase the number of quantization levels, and in the quantization of the color signal U, the range of the color signal U is increased. Is divided into a wider and smaller number of sections to make the number of quantization levels smaller than the number of quantization levels of the color signal V. By adopting such a method, it is possible to make the color error caused by the quantization of the color signal uniform and make the quality of the image obtained after decoding (after expansion) the color signal uniform.

  In the above description, the Lab color difference can be a color difference in the CIE Lab space. In general, the Lab color difference quantitatively indicates the color difference perceived by the human eye and relatively reflects the human psychological quantity with respect to the color difference, and is calculated by simple calculation. . For this reason, according to the quantization according to the present embodiment as described above, it is possible to achieve quantization that reflects human perceptual characteristics relatively well.

  Here, the Lab color difference is characterized by being a simple and generally usable index. However, in order to more accurately reflect the human psychological amount with respect to the color difference, the Lab color difference is an amount obtained by correcting the Lab color difference. It is preferable to use some modified CIELAB color difference.

  Hereinafter, an example of image processing by the quantization method will be described more specifically. FIG. 8 is a block diagram illustrating an image processing method according to Embodiment 1 of the present invention. As illustrated in FIG. 8, the CPU 11 illustrated in FIGS. 1 and 2 includes a 2 × 2 pixel cutout unit 33, an RCT unit 35, a Y orthogonal conversion unit 37, a U averaging unit 39, and a V average. A quantization unit 41, a Y quantization unit 43, a U quantization unit 45, and a V quantization unit 47.

  Here, the 2 × 2 pixel cutout unit 33 is supplied with an image signal having an RGB component having a value of 0 to 255 (8 bits), and the R component, the G component, and the B component are cut out every four pixels. , And supplied to the RCT unit 35. Then, the RCT unit 35 performs RCT on the RGB components in units of four pixels, and converts the converted signal Y to the Y orthogonal transformation unit 37, the color signal U to the U averaging unit 39, and the color in units of four pixels. The signal V is supplied to the V averaging unit 41, respectively.

  Further, the Y orthogonal transform unit 37 orthogonally transforms the supplied signal by discrete cosine transform or Haar transform and supplies the signal to the Y quantization unit 43. Further, the U averaging unit 39 obtains an average value Ua of the supplied signal and supplies it to the U quantizing unit 45. Further, the V averaging unit 41 obtains the average value Va of the supplied signal and supplies it to the V quantization unit 47. Here, in the above, instead of obtaining an average value in the color signals U and V, orthogonal transformation similar to that of the signal Y may be performed. In this case, the quantization method according to the present invention can be applied to the low-frequency component after the orthogonal transformation.

  The orthogonally transformed signal Y is quantized (encoded) by the Y quantizing unit 43, and the average value Ua obtained in units of 4 pixels is quantized (encoded) by the U quantizing unit 45. The average value Va obtained for each pixel is quantized (encoded) by the V quantization unit 47.

  Here, the quantization operation of the color signal V by each unit shown in FIG. 8 will be described with reference to the flowchart shown in FIG. First, in step S1, 8-bit color signal V is input for 4 pixels of 2 × 2 in length and width. Next, in step S2, an average value Va of the color signals V for the four pixels is calculated.

  In step S3, it is determined whether or not the calculated average value Va is 0 or more. If it is 0 or more, the process proceeds to step S4. If it is less than 0, the process proceeds to step S10. Next, in step S4, the average value Va indicated by 8 bits is linearly quantized to 7-bit data by dividing by 2, and in step S10, an 8-bit average value taking values from 0 to 255 is obtained. -Va (| Va |) is nonlinearly quantized into 6-bit data having a quantization level number half of 7 bits with reference to the table defining the quantization threshold shown in Table 2 below. .

ここで、平均値Ｖａの単位量子化誤差当りに生じるＬａｂ色差は、概ね原点０付近で最大となるため、上記のように平均値Ｖａの極性（正負）に応じて量子化レベル数が変えられる。

Here, since the Lab color difference generated per unit quantization error of the average value Va is approximately the maximum near the origin 0, the number of quantization levels can be changed according to the polarity (positive / negative) of the average value Va as described above. .

  In step S5, a 1-bit code is added to the average value | Va | quantized in step S4 or S10 as a prefix bit in order to determine the polarity when the color signal V is decoded. Ends the operation.

  Next, the quantization operation of the color signal U by each unit shown in FIG. 8 will be described with reference to the flowchart shown in FIG. First, in step S1, 8-bit color signal U is input for 4 pixels of 2 × 2 in length and width. Next, in step S2, an average value Ua of the color signals U for the four pixels is calculated.

  In step S3, it is determined whether or not the calculated average value Ua is 0 or more. If it is 0 or more, the process proceeds to step S4, and if it is less than 0, the process proceeds to step S10. Next, in step S4, the average value Ua indicated by 8 bits is linearly quantized to 6-bit data by dividing by 4, and in step S10, an 8-bit average value taking values from 0 to 255 is obtained. -Ua (| Ua |) is nonlinearly quantized into 5-bit data with reference to a table defining quantization thresholds.

  Here, since the Lab color difference generated per unit quantization error of the average value Ua is approximately the maximum near the origin 0, the number of quantization levels can be changed according to the polarity (positive / negative) of the average value Ua as described above. . As described above, the number of quantization levels of the average value Ua is made smaller than the number of quantization levels of the average value Va regardless of whether the average value Ua is positive or negative.

  In step S5, a 1-bit code is added to the average value | Ua | quantized in step S4 or S10 as a prefix bit in order to determine the polarity when the color signal U is decoded. Ends the operation.

  Next, an example of the Lab color difference calculation method will be described more specifically. FIG. 11 is a block diagram illustrating a Lab color difference calculation method according to Embodiment 1 of the present invention. As illustrated in FIG. 11, the CPU 11 illustrated in FIGS. 1 and 2 includes Lab conversion units 51 and 53, an input profile unit 49, and a Lab color difference calculation unit 55.

  Here, the Lab conversion units 51 and 53 are supplied with information defining the conversion from the input profile unit 49. The “input profile” means a table for converting RGB components of an image signal into Lab components. Then, the Lab conversion units 51 and 53 convert the supplied RGB components into Lab components by using the table.

  As an example, when the data D1 including the R component, the G component, and the B component is supplied to the Lab conversion unit 51, and the data D2 including the R ′ component, the G component, and the B component is supplied to the Lab conversion unit 53, As shown in FIG. 11, data D1 is converted into an L component, an a component, and a b component, and data D2 is converted into an L ′ component, an a ′ component, and a b ′ component. The L component, the a component, the b component, the L ′ component, the a ′ component, and the b ′ component are all supplied to the Lab color difference calculation unit 55.

At this time, the Lab color difference calculation unit 55 calculates the color difference ΔE ^* in the Lab space by the following equation.

ΔE ^* = {(L−L ′) ² + (a−a ′) ² + (b−b ′) ² } ^{1/2 Although} the Lab color difference can be easily obtained, In order to ensure accuracy by dealing with the above, it is preferable to use a modified Lab color difference ΔEm. Here, the corrected Lab color difference ΔEm is calculated by the following equation when h is 0.75, l is 1.00, and c is 0.50.

ΔEm = {(hΔH ^* ) ² + (1ΔL ^* ) ² + (cΔC ^* ) ² } ^{1/2 where} ΔH ^* = {ΔE ^{* 2−} ΔL ^{* 2−} ΔC ^{* 2} } ^1/2 ΔL ^* = L−L′ΔC ^* = {a ² + b ² } ^1/2 − {a ′ ² + b ′ ² } ^1/2

  From the above, according to the image processing apparatus and the image processing method according to Embodiment 1 of the present invention, the color difference per unit quantization error of the color signal is maximized, or before and after the polarity change point of the color signal. Since the number of quantization levels of the color signal is changed, quantization considering the asymmetry of the color difference with respect to the color signal is realized to improve the quantization (encoding) efficiency, and the color error with respect to the original image is reduced. Small images can be generated.

[Embodiment 2]
As shown in FIG. 6, since the Lab color difference depends on the value of the color signal V, in practice, it is difficult to accurately determine the number of quantization levels of the color signal U. In view of this, it is conceivable to change the quantization method of the color signal in two regions divided by the magnitude relationship between the color signal U and the color signal V on the UV plane shown in FIG.

  That is, in the second embodiment, Lab is changed by changing the quantization method (quantization level number, quantization threshold value, etc.) of the color signal to be quantized according to the color signal not to be quantized. Color difference is suppressed. More specifically, the color difference per unit quantization error of the color signal to be quantized is represented by a region on the plane constituted by the color signal not to be quantized and the color signal to be quantized. Is divided by a locus of points where the maximum value is (straight line V = U in the example shown in FIG. 6), and the quantization method is changed in each divided region.

  Further, when the color signal V is larger than the color signal U, the color signal U is nonlinearly quantized while referring to a table storing quantization thresholds, and when the color signal V is smaller than or equal to the color signal U, the color is displayed. The signal U can be linearly quantized. If quantization is performed by such a method, when the color signal V has a magnitude equal to or smaller than the color signal U, the quantization can be performed without using the table as described above. The amount of memory required can be reduced.

  Note that the linear quantization is preferably applied in a range where the color difference per unit quantization error of the color signal is a constant value, as in the first embodiment.

  Also in the second embodiment, if the number of quantization levels of the color signal V is larger than the number of quantization levels of the color signal U as in the first embodiment, the color difference caused by the quantization error is made uniform. As a result, the image quality after decoding (after decompression) can be made uniform.

  The image processing apparatus according to the second embodiment has the same configuration as that of the image processing apparatus according to the first embodiment shown in FIG. 1, and the CPU 11 includes each unit shown in FIG. Hereinafter, a method of quantizing the color signal U according to the second embodiment by the CPU 11 will be described with reference to the flowchart of FIG.

  First, in step S1, 8-bit color signal U is input for 4 pixels of 2 × 2 in length and width. Next, in step S2, the U averaging unit 39 calculates an average value Ua of the color signals U for the four pixels.

  In step S3, it is determined whether or not the calculated average value Ua is equal to or greater than the average value Va calculated by the V averaging unit 41. If the average value Ua is equal to or greater than the average value Va, step S4 is performed. If the average value Ua is less than the average value Va, the process proceeds to step S10. Next, in Step S4, since the Lab color difference generated in the quantization of the color signal U is large and substantially constant, the average value Ua indicated by 9-bit data including 1 bit indicating positive / negative is changed to a positive / negative sign. The 8-bit data indicating the absolute value is divided by 4 and the linear quantization is performed to 6-bit data.

  On the other hand, in Step S10, since the Lab color difference generated in the quantization of the color signal U is small and non-linear, the quantization threshold is maintained while maintaining a positive / negative sign with respect to the average value Ua indicated by the 9-bit data. The 8-bit data indicating the absolute value is nonlinearly quantized into 5-bit data while referring to a table that defines the values.

  In step S5, a 1-bit code is added as a prefix bit to the average value Ua quantized in step S4 or step S10 in order to determine the magnitude of the color signal U and the color signal V at the time of decoding. The quantization operation is finished.

  Other operations such as quantization of the signal Y and the average value Va and calculation of the Lab color difference are performed in the same manner as the image processing apparatus according to the first embodiment.

  As described above, according to the image processing apparatus and the image processing method according to the second embodiment of the present invention, the quantization method of the color signal to be quantized is changed according to the color signal not to be quantized. By performing the quantization in consideration of the characteristics of the color signals that have been performed, the quantization efficiency can be increased and the color difference caused by the quantization can be suppressed.

[Embodiment 3]
In the above embodiment, the method of quantizing (encoding) the color signal U and the color signal V has been described. However, if the characteristics of the Lab color difference generated in the quantization of the color signal U and the color signal V are taken into consideration, UV It is also conceivable to quantize the signed distance L from the straight line V = U on the plane and the color signal V.

  That is, as shown in FIG. 13, the color difference caused by the quantization error of the color signal U substantially depends on the distance from the straight line V = U, and does not depend on the values of the color signals U and V themselves. Therefore, if the distance from the straight line V = U is quantized with a positive or negative sign, the number of quantization levels and the quantization threshold can be determined in consideration of the characteristics of the Lab color difference.

  If the signed distance L from the quantized straight line V = U and the color signal V are decoded, the color signals U and V themselves can be decoded, and the RGB components can be decoded. Such a quantization method is preferably applied to RCT, but can also be applied to other color conversions.

As described above, when the signed distance L from the straight line V = U is quantized, it is complicated to obtain the distance itself and encode it. It is preferable to quantize the quantity. Therefore, for example, as shown in FIG. 13, when a point P (u, v) on the UV plane is considered, a U-intercept of a straight line V = U−u + v having a slope 1 passing through the point P is a point Q (u−v , 0), it can be seen that the value of (u−v), which is the U coordinate of this point Q, is 2 ^1/2 times the signed distance L. Therefore, in this embodiment, the difference between the color signal U and the color signal V is quantized.

  Note that, also in the quantization method according to the present embodiment, when the difference between the color signal U and the color signal V is less than 0 as in the quantization method according to the above embodiment, the quantization threshold value is stored. The color signal U can be nonlinearly quantized with reference to the table, and when the difference between the color signal U and the color signal V is 0 or more, the color signal U can be linearly quantized.

  Further, since the Lab color difference caused by the quantization error of the color signal V takes a larger value as a whole than the Lab color difference caused by the quantization error of the color signal U, it is the same as the quantization method according to the above embodiment. In addition, it is preferable that the number of quantization levels of the color signal V is larger than the number of quantization levels for the difference between the color signal U and the color signal V.

  The image processing apparatus according to the third embodiment has the same configuration as the image processing apparatus according to the first embodiment shown in FIG. 1, and the CPU according to the present embodiment corresponding to the CPU 11 includes: Each part shown in FIG. 14 is included. That is, as shown in FIG. 14, the CPU according to the present embodiment includes a 2 × 2 pixel cutout unit 33, an RCT unit 35, a Y orthogonal conversion unit 37, and a (U−V) averaging unit 57. , V averaging unit 41, Y quantization unit 43, (U−V) quantization unit 59, and V quantization unit 47.

  The CPU having the above-described configuration operates in the same manner as the CPU 11 according to the above embodiment. However, the RCT unit 35 performs RCT on the RGB components in units of four pixels, and converts the signals in units of four pixels. Y is supplied to the Y orthogonal transform unit 37, the color signal U and the color signal V are supplied to the (U−V) averaging unit 57, and the color signal V is supplied to the V averaging unit 41.

  Then, the (U−V) averaging unit 57 obtains the average value (U−V) a of the difference between the color signal U and the color signal V in units of four pixels according to the supplied color signal U and color signal V. , (U-V) is supplied to the quantization unit 59. The average value (U−V) a is quantized (encoded) by the (U−V) quantization unit 59. Note that, by the quantization, the value of the average value (U−V) a is expressed by 9-bit data including a 1-bit sign indicating positive / negative and 8-bit data indicating absolute value.

  In the following, the quantization operation of the signal (U-V) by each unit shown in FIG. 14 will be described with reference to the flowchart of FIG. First, in step S1, 8-bit color signal U is input for 4 pixels of 2 × 2 in length and width. Next, in step S2, an average value (U−V) a of the signals (U−V) for the four pixels is calculated.

  In step S3, it is determined whether or not the calculated average value (U−V) a is 0 or more. If it is 0 or more, the process proceeds to step S4. If it is less than 0, step S10 is performed. Proceed to Next, in step S4, the average value (U−V) a whose absolute value is represented by 8 bits is divided by 4 while retaining the sign indicating the sign of the average value (U−V) a. Linear quantization into 6-bit data.

  On the other hand, in step S10, an 8-bit absolute value that takes a value from 0 to 255 is referred to a table that defines the quantization threshold while retaining the sign indicating the sign of the average value (U−V) a. Then, nonlinear quantization is performed to 5-bit data.

  Here, since the Lab color difference generated per unit quantization error of the average value (U−V) a is approximately the maximum near the origin 0, the polarity (positive / negative) of the average value (U−V) a as described above. The number of quantization levels can be changed according to. Further, as described above, the number of quantization levels of the average value (U−V) a is made smaller than the number of quantization levels of the average value Va regardless of whether the average value (U−V) a is positive or negative.

  In step S5, a 1-bit code is prefixed to the average value (U−V) a quantized in step S4 or step S10 in order to determine whether the color signal U is positive or negative (polarity). The bit is added and the quantization operation is completed.

  Other operations such as the quantization of the signal Y and the average value Va and the calculation of the Lab color difference are performed in the same manner as the image processing apparatus according to the above embodiment.

  As described above, according to the image processing apparatus and the image processing method according to Embodiment 3 of the present invention, the color signal U and V are quantized with the signed distance from the straight line V = U and the color signal V in the UV plane. Therefore, it is possible to realize quantization by taking into account the characteristics of the color signals after quantization by a simple method, increase the quantization efficiency, and suppress the color difference caused by the quantization.

  The image processing method according to the embodiment of the present invention described above can be described by a computer program. Accordingly, the program describing the image processing method is recorded on a recording medium such as a CD-ROM or a floppy (registered trademark) disk, and the recording medium is mounted on a computer to execute the program. The method can be easily realized.

  Therefore, the image processing method according to the present embodiment can be widely applied to devices that handle color images, including various application programs and device drivers such as printer drivers.

It is a figure which shows the structure and operation | movement of an image processing apparatus which concerns on Embodiment 1 of this invention. It is a figure explaining other operation | movement of the image processing apparatus shown by FIG. It is a 1st figure explaining the method of calculating | requiring the Lab color difference per unit quantization error. It is a 2nd figure explaining the method of calculating | requiring the Lab color difference per unit quantization error. FIG. 6 is a diagram illustrating a color difference caused by a quantization error of a color signal V. FIG. 4 is a diagram illustrating a level of color difference caused by a quantization error of a color signal U. 4 is a graph showing Lab color difference per unit quantization error caused by a quantization error of a color signal U. It is a block diagram explaining the image processing method which concerns on Embodiment 1 of this invention. It is a flowchart which shows the encoding method of the color signal V which concerns on Embodiment 1 of this invention. It is a flowchart which shows the encoding method of the color signal U which concerns on Embodiment 1 of this invention. It is a block diagram explaining the calculation method of the Lab color difference which concerns on Embodiment 1 of this invention. It is a flowchart explaining the encoding method of the color signal U which concerns on Embodiment 2 of this invention. It is a figure explaining the image processing method which concerns on Embodiment 3 of this invention. It is a block diagram explaining the image processing method which concerns on Embodiment 3 of this invention. It is a flowchart explaining the encoding method of the color signal U which concerns on Embodiment 3 of this invention.

Explanation of symbols

1 Hard disk 3, 7, 19 First storage area 4, 9, 21 Second storage area 5, 17 Random access memory (RAM)
11, 15 Central processing unit (CPU)
13 Printer 23 Envelopes 25 and 32 Curves 26 to 31 Region 33 2 × 2 pixel cutout unit 35 RCT (Reversible Component Transform) unit 37 Y orthogonal transform unit 39 U averaging unit 41 V averaging unit 43 Y quantization unit 45 U Quantization unit 47 V quantization unit 49 Input profile unit 51, 53 Lab conversion unit 55 Lab color difference calculation unit 57 (UV) averaging unit 59 (UV) quantization unit BS Data bus PC Personal computer

Claims (2)

  An image processing method for processing a color component extracted from an image signal, wherein a color difference per unit error of the color component generated when the color component is quantized is determined as a G component in an RGB component of the image signal. And a step of quantizing the color component by dividing the value of the color component into a plurality of ranges for each range in which the color difference is equal to or less than a predetermined value. Image processing method.   An image processing method for processing a color component extracted from an image signal, wherein a color difference per unit error of the color component generated when the color component is quantized is determined as a G component in an RGB component of the image signal. And a step of quantizing the color component by dividing the value of the color component into a plurality of ranges for each range in which the color difference is equal to or less than a predetermined value. An image processing method.

Images